Re: [PATCH] virtio_blk: add block topology support

[Rusty Russell <rusty@rustcorp.com.au>]

On Fri, 15 Jul 2011 03:36:42 +0400, Alexey Zaytsev <alexey.zaytsev@gmail.com> wrote:
> On Mon, Feb 1, 2010 at 04:10, Rusty Russell <rusty@rustcorp.com.au> wrote:
> > On Sun, 31 Jan 2010 06:49:10 am Christoph Hellwig wrote:
> >> On Sat, Jan 30, 2010 at 03:29:49PM +1030, Rusty Russell wrote:
> >> > I bow to your expertise on that.  My only query is the __u16 for min_io_size; is that likely to restrict us?
> >>
> >> Looks like you caught me there - I wrote the above odd format about the
> >> physical_block exponent, but scsi actually does the min_io and opt_io
> >> size in logical blocks, too.  With that in account the u16 as in scsi
> >> is perfectly fine.
> >
> > Thanks, applied.
> 
> Ugh, guys. I know it's already applied long ago, but this kind of
> contradicts the virtio specification, doesn't it?

Ugh indeed!  The same field is used by two places, as
VIRTIO_BLK_F_SECTOR_MAX never made it into the linux headers.

> VIRTIO_BLK_F_SECTOR_MAX (10) Maximum total sectors in
>    an I/O.

The patch made the spec, but as far as I can tell, no implementation.

If that's right, we just remove it from the spec.  If it did make it in,
we now have a very ugly case where the layout will have to vary
depending on what options are negotiated.

Once we've resolved this, I'll update the spec...

Thanks,
Rusty.